Data augmentation using synthetic data for time series classification with deep residual networks

Fawaz, Hassan Ismail; Forestier, Germain; Weber, Jonathan; Idoumghar, Lhassane; Müller, Pierre-Alain

doi:10.48550/arxiv.1808.02455

Cited by 21 publications

(25 citation statements)

References 12 publications

(22 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In order to visualize the input data, we use histogram plots. For example, the input maximum feret is depicted in Figures 8 and 9 In comparison to that we obtain a bad fit in the cases where a trained SimResNet of the data set V (RN) is applied on the data set RN (V) (see Figures 9,11). The reason can be obtained in the different shapes of the histograms of the input on each data set which can be clearly deduced in the Fig.…”

Section: Single Featurementioning

confidence: 80%

See 1 more Smart Citation

Simplified ResNet approach for data driven prediction of microstructure-fatigue relationship

Gebhardt¹,

Trimborn²,

Weber³

et al. 2020

Preprint

View full text Add to dashboard Cite

The heterogeneous microstructure in metallic components results in locally varying fatigue strength. Metal fatigue strongly depends on size and shape of non-metallic inclusions and pores, commonly referred to as "defects". Nodular cast iron (NCI) contains graphite inclusions (nodules) whose shape and frequency influence the fatigue strength. Fatigue strength can be simulated by micromechanical finite element models. The drawback of these models are the large computational costs. Therefore, we employ a data-driven machine learning methodology. More precisely, we utilize the simplified residual neural network (SimResNet) which was recently introduced in [16] to predict fatigue strength from metallographic data.For the training, we use fatigue data which is simulated with a micromechanical model and the shakedown theorem. The micromechanical models are derived directly from micrographs of nodular cast iron, respectively. The application of SimResNet shows a good performance to predict fatigue strength by local microstructures of nodular cast iron. We show several test cases. The simplified character of SimResNet enables fast predictions of fatigue by microstructures, even in comparision to classical residual neural networks.

show abstract

Section: Single Featurementioning

confidence: 80%

“…ResNet have been successfully applied to a variety of applications such as image recognition [48], robotics [50] or classification [11]. More recently, also applications to mathematical problems in numerical analysis [32,33,45] and optimal control [35] have been studied.…”

Section: Introductionmentioning

confidence: 99%

Simplified ResNet approach for data driven prediction of microstructure-fatigue relationship

Gebhardt¹,

Trimborn²,

Weber³

et al. 2020

Preprint

View full text Add to dashboard Cite

show abstract

“…We believe adding more data to the training dataset will further improve the performance of AdaLSTM especially for the data from the sensor on the wrists and the arms of the users. We are planning to address this issue in two directions: (1) conduct an extensive multi-modality data collection from a large number of participants performing different lying postures, including main postures and their other variations; (2) produce signal-/feature-level synthesis data using data augmentation techniques such as rotation, permutation, time-wrapping, scaling, magnitude-wrapping, jittering [39], sequence to sequence learning techniques [40], and generative adversarial networks [41].…”

Section: Discussionmentioning

confidence: 99%

Pervasive Lying Posture Tracking

Alinia¹,

Samadani²,

Milošević³

et al. 2020

Preprint

View full text Add to dashboard Cite

Automated lying-posture tracking is important in preventing bed-related disorders such as pressure injuries, sleep apnea, and lower-back pain. Prior research studied in-bed lying posture tracking using sensors of different modalities (e.g., accelerometer and pressure sensors). However, there remain significant gaps in research about how to design efficient in-bed lying posture tracking systems. These gaps can be articulated through several research questions as follows. First, can we design a single-sensor, pervasive, and inexpensive system that can accurately detect lying postures? Second, what computational models are most effective in the accurate detection of lying postures? Finally, what physical configuration of the sensor system is most effective for lying posture tracking? To answer these important research questions, in this article, we propose a comprehensive approach to design a sensor system that uses a single accelerometer along with machine learning algorithms for in-bed lying posture classification. We design two categories of machine learning algorithms based on deep learning and traditional classification with handcrafted features to detect lying postures. We also investigate what wearing sites are most effective in accurate detection of lying postures. We extensively evaluate the performance of the proposed algorithms on nine different body locations and four human lying postures using two datasets. Our results show that a system with a single accelerometer can be used with either deep learning or traditional classifiers to accurately detect lying postures. The best models in our approach achieve an F-Score that ranges from 95.2% to 97.8% with 0.03 to 0.05 coefficient of variation. The results also identify the thighs and chest as the most salient body sites for lying posture tracking. Our findings in this article suggest that because accelerometers are ubiquitous and inexpensive sensors, they can be a viable source of information for pervasive monitoring of in-bed postures.

show abstract

“…Data augmentation can create several new feature spaces and increase the amount of training data without additional ground truth labels, which has been widely used to improve the performance and generalizability of downstream predictive models. Many works have proposed data augmentation technologies on different types of features, such as images [7,28,15], texts [12,29], vectorized features [18,6], etc. However, how to effectively augment graph data remain a challenging problem, as graph data is more complex and has non-Euclidean structures.…”

Section: Introductionmentioning

confidence: 99%

“…Many prior studies [6,21] in data augmentation are to capture the interactions between features by taking addition, subtraction, or cross product of two original features, which are suitable for tensorial features. The major obstacle in graph data is that the original features, graph topology, and node attributes, are two types of data, one is usually encoded by position in Euclidean space, while the other is encoded by node connectivity in non-Euclidean space.…”

Section: Introductionmentioning

confidence: 99%

Data Augmentation for Graph Convolutional Network on Semi-Supervised Classification

Tang

Qiao

Hong

et al. 2021

Preprint

View full text Add to dashboard Cite

Data augmentation aims to generate new and synthetic features from the original data, which can identify a better representation of data and improve the performance and generalizability of downstream tasks. However, data augmentation for graph-based models remains a challenging problem, as graph data is more complex than traditional data, which consists of two features with different properties: graph topology and node attributes. In this paper, we study the problem of graph data augmentation for Graph Convolutional Network (GCN) in the context of improving the node embeddings for semi-supervised node classification. Specifically, we conduct cosine similarity based cross operation on the original features to create new graph features, including new node attributes and new graph topologies, and we combine them as new pairwise inputs for specific GCNs. Then, we propose an attentional integrating model to weighted sum the hidden node embeddings encoded by these GCNs into the final node embeddings. We also conduct a disparity constraint on these hidden node embeddings when training to ensure that non-redundant information is captured from different features. Experimental results on five real-world datasets show that our method improves the classification accuracy with a clear margin(+2.5% -+84.2%) than the original GCN model.

show abstract

Data augmentation using synthetic data for time series classification with deep residual networks

Cited by 21 publications

References 12 publications

Simplified ResNet approach for data driven prediction of microstructure-fatigue relationship

Simplified ResNet approach for data driven prediction of microstructure-fatigue relationship

Pervasive Lying Posture Tracking

Data Augmentation for Graph Convolutional Network on Semi-Supervised Classification

Contact Info

Product

Resources

About